Slashdot Mirror
Ask Slashdot: How Do You Assess the Status of an Open Source Project?
Chrisq writes: "Our software landscape includes a number of open source components, and we currently assume that these components will follow the same life-cycle as commercial products: they will have a beta or test phase, a supported phase, and finally reach the end of life. In fact, a clear statement that support is ended is unusual. The statement by Apache that Struts 1 has reached end of life is almost unique. What we usually find is:
- Projects that appear to be obviously inactive, having had no updates for years
- Projects that are obviously not going to be used in any new deployments because the standard language, library, or platform now has the capability built in
- Projects that are rapidly losing developers to some more-trendy alternative project
- Projects whose status is unclear, with some releases and statements in the forums that they are 'definitely alive,' but which seem to have lost direction or momentum.
- Projects that have had no updates but are highly stable and do what is necessary, but are risky because they may not interoperate with future upgrades to other components.
By the treating Open Source in the same way as commercial software we only start registering risks when there is an official announcement. We have no metric we can use to accurately gauge the state of an open source component — but there are a number of components that we have a 'bad feeling' about. Are there any standard ways of assessing the status of an open source project? Do you use the same stages for open source as commercial components? How do you incorporate these in a software landscape to indicate at-risk components and dependencies?"
This isn't a problem that is unique to open source. Several commercial libraries that we have used in the past have entered the twilight zone where the developer is neglecting them, and refuses to release any sort of roadmap or EOL announcement. Eventually, you just have to make your own call based on how much work it will be to move to a new library vs the risk of staying with the current one. At least with open source if you get stuck with a dead library you can choose to take over maintaining it on your own either as a long term strategy or a short-term stop-gap until you can move onto something else.
One metric yielding interesting results is the concept of "technical debt", as introduced by Martin Fowler. Sonar Source, for example, measures this metric very well. A project that has seen neither increase ( recently taken risk ) nor decrease ( recent moves toward stabilization ) may very well be dead. I recently used it upon our own software of 580 KSLOC. The interesting conclusion: core stable, some utilities half dead or worse, much life springing up at the functional fringes. This also holds for e.g tomcat. The tactical and strategical conclusions one may draw from such considerations are fascinating.
Religous speak to God. Insane are spoken to by God. When all shut up, one can finally hear Shostakovich in peace
Yeah you want to be careful with activity metrics. Awk hasn't seen many updates in the last two years. Mostly because it hasn't NEEDED much in the last ten or twenty years. That means it's already rock solid, not that it should be avoided.
0) If the project does what you need today, USE IT. Don't get so bound up in "future-proofing" your technology stack that you get paralyzed looking for "the perfect product that will do exactly what we want forever and never let us down."
1) Define your standard software stack. Mandate that all software written internally using open source components use these standard components & versions, or coordinate making a new version available to all projects if there's a particular new feature of a new version that is absolutely mandatory;
2) Always, always, always, download source for the version of the package you're installing (even if you just grab binary-only distributions to install & run), and archive it for posterity in some location YOU control and backup - DO NOT rely on "the internet" to help you find an old version of software; this allows you to fix (or hire someone to fix) any problem you have down the road in case of real critical issues where no active project maintainers can be found/hired/worked with.
3) Every few months (we shoot for ~6 months), review your stack and grab the latest versions of each component and make it available in your dev / testing environments;
4) If a component starts getting stale (no updates for 2 or more of these cycles), we'll start thinking about replacements for that component, and investigate likely alternatives, and bump this item up into the "needs monitoring" risk category - no production impact yet, but as soon as you need to release a patch of that production version using the outdated component, you're gonna be in trouble.
5) Periodically (nightly if you have resources - get something like jenkins or similar for this sort of thing) ensure that you can build these components from source successfully. Especially as they get 'stale,' you'll run into issues - system libs, headers, etc. will change over time, and there will come a point where you are no longer able to build the software without code modification. At that point, if any of your software is still using the version, then you should start raising alarms and bump the risk level up to "severe." This could cripple your production env.
6) If a crisis comes up and a dead project is the culprit... well, we've got the code and can always modify it ourselves, if we haven't found any suitable alternative.
There's really no magic to it - just make sure that developers aren't downloading "every version under the sun," and ensure that the versions you're using are reproducible, available, and actively managed on your end. Risk management is paramount.